Erik Supit
31.7K posts

Erik Supit
@eriksupit
Jakarta | Building https://t.co/0735bUZ3Sw — an AI Agent platform for academic research and paper generation.



Holy shit… this study just flipped everything we thought about AI on its head 🤯 Penn State researchers discovered something *wild*: being rude to AI actually makes it smarter. They tested ChatGPT-4o with 250 questions across 5 politeness levels: • Very polite → 80.8% accuracy • Polite → 81.4% • Neutral → 82.2% • Rude → 82.8% • 'Very rude → 84.8%' Statistical tests confirmed this isn’t random harsh prompts consistently outperformed polite ones. Even crazier: older models like GPT-3.5 behaved the *opposite* way. But GPT-4 and newer? They get sharper when you push them. Paper: arxiv. org/abs/2510.04950








